AI for accountants: what to automate first in 2026

Accountants get the most from AI on the admin around client work, not on the judgement: chasing clients for records, reading receipts and bank lines into the ledger, answering routine client questions in busy season, and finding answers in the firm's own procedures. Start with what your software already includes, because QuickBooks and Xero both now use AI to categorise and match transactions, then connect the gaps between your practice management tool, your ledger and your inbox. Keep a person on advice, judgement calls and anything filed with HMRC or the IRS. Below are the eight jobs worth handing over, what the software does on its own, the client-data rules to settle first, and a sample project showing how we run one and what to expect.

How are accounting firms using AI in 2026?

Widely, but not yet well. Karbon's 2026 State of AI in Accounting report, based on nearly 600 accounting professionals, found that 98% of firms use AI, saving an average of an hour a day per employee, yet only 21% have an AI policy or strategy. A Financial Cents survey of 486 accounting and bookkeeping professionals in North America, published in September 2026, found the same pattern: 95% of firms are using or exploring AI, but only about one in five report a clear, measurable return, and 96% of users rely mainly on general-purpose chatbots.

The difference is how AI is used. Firms that have built it into their daily workflows are far more likely to see a return than those using it now and then:

Chart

Firms that report a measurable return from AI, by how they use it

• Running: AI built into daily workflows56%

  11% of firms. A small group (52 people), so read it as a direction, not a precise figure.

• Walking: used regularly for some jobs24%

  36% of firms.

• Crawling: occasional, individual use4%

  31% of firms.

Show the data as a table

Stage	Share of firms	Report measurable ROI
Running	11%	56%
Walking	36%	24%
Crawling	31%	4%
Exploring (not yet in daily use)	17%	n/a
Not using AI	5%	n/a

The lesson for a small or mid-sized practice is not "use more AI". It is to connect AI to the work that already flows through your firm, so the time it saves is real and measurable, rather than asking a chatbot the odd question.

8 jobs worth handing to AI in an accounting practice

If you are not sure which of these to start with, our method for choosing what to automate first scores each job on value and effort, and our AI agent examples show how the multi-step ones work.

What QuickBooks and Xero now do themselves

The ledgers have moved quickly this year, and the first step for any firm is to use what is already included.

QuickBooks (US). Intuit is discontinuing QuickBooks Online Accountant on 31 December 2026 and moving firms to Intuit Accountant Suite, where its AI agents sit, and whose Core plan is free. Billing for its paid features begins on 20 January 2027, and firms trying them in beta that take no action by 19 January drop back to Core; the Books Close add-on is listed at $8 per client a month for firms with 50 clients or fewer, and $6 for larger firms. If you run a US practice on QuickBooks, plan that move before tax season, not during it.

Xero. At Xerocon London in July 2026, Xero described its AI assistant, JAX, as reading source documents and extracting the data into Xero, and matching transactions to bank feeds automatically. It said document chasing for missing receipts is coming soon.

What the ledgers don't do is the work between systems: the client who emails a photo instead of using the portal, the reminder that should come from your practice management tool, the question asked in a phone call, the procedure that lives in a shared drive. That is where most of a practice's admin time goes, and it is where a done-for-you setup earns its place.

Sample project: how we work, and what to expect

Here is how we would run a project for a typical practice. The practice is fictional, a composite of the firm size we work with most, and every figure below is an assumption you can swap for your own.

WORK ORDER · SAMPLE PROJECT6 STAFF · 300 CLIENTS

AI FOR AN ACCOUNTING PRACTICE

A fictional practice, shown to explain how we run these projects. Your timings and steps will differ.

1. 01 · DAY 1

   Free 20-minute call

   We ask what eats your week, which software you run and how many clients you have.

   YOU SEE: A plain answer on whether AI is worth it for you, and where to start.

2. 02 · WEEK 1

   Map and measure

   We read a sample of last quarter's chase emails and client questions, and time the jobs with your team.

   YOU SEE: A written quote: what we will build, what it costs, and the hours it should save, with the assumptions shown.

3. 03 · WEEKS 2-3

   Build, in draft mode

   Document chasing linked to your practice management tool, receipt capture into the ledger with a review queue, drafted replies to routine client questions.

   YOU SEE: Everything prepares work for your team; nothing is sent or posted on its own.

4. 04 · WEEK 4

   Test on real cases

   Your team runs last quarter's cases through it. We fix what it gets wrong and agree the client-consent and data steps.

   YOU SEE: A list of what it handles alone, what it drafts, and what always stays with you.

5. 05 · WEEKS 5-8

   Go live, gradually

   Automations switched on job by job, with a weekly report against the baseline from week 1.

   YOU SEE: Hours spent chasing and keying, before and after, in plain numbers.

6. 06 · ONGOING

   Look after

   Monitoring, fixes when your software changes, and a quarterly review before each busy period.

   YOU SEE: A system that keeps working through tax season, and someone to call when it doesn't.

NOTHING STARTS UNTIL YOU SAY YESQUOTED IN WRITING

Each stage matches how we work on every project: a free call, a written quote before anything starts, a build you test yourself, and someone looking after it afterwards.

The arithmetic, with the assumptions shown

For a six-person practice with 300 clients, of whom 40 have their books done monthly:

About 26 hours a month all year, and around 38 in tax season: the best part of a week of one person's time every month, given back to client work. Your numbers will differ, which is why the first week of any project is spent measuring them.

The outcomes worth expecting

• Hours back, measured against a baseline taken before the build, not estimated afterwards.
• Records arrive earlier, because reminders go out on time, every time, so less of the work piles up at the deadline.
• Clients get answers at any hour in busy season, and your team stops repeating the same five replies.
• A trail you can show: every reminder, capture and draft is logged, and nothing is sent or posted without a person's say-so.
• A written AI policy, which most firms still don't have: only 21% in Karbon's survey.

And what not to expect: AI won't give advice, sign anything off or replace anyone's judgement, and it won't be finished in a week. A first set of automations typically takes four to eight weeks to build, test and switch on carefully.

When you are ready to see what this would look like for your firm, you can book a free 20-minute call or build a quote in a couple of minutes.

Client data: the rules to settle first

United States. Section 7216 of the Internal Revenue Code makes it a misdemeanour for anyone who prepares tax returns to knowingly or recklessly disclose, or use, information given to them for a return, other than to prepare it, with fines and up to a year in prison. As of October 2026, the IRS has not said whether sending return information to an AI tool counts as a disclosure, so the cautious course is to get signed client consent in the IRS's required format before tax data goes into one, use business plans that don't train on your data, and keep a non-AI route for clients who decline.

United Kingdom. UK GDPR applies to client data, and your professional body's code (ICAEW, ACCA or AAT) sets confidentiality duties. Use AI providers under a data processing agreement, tell clients how their data is handled, and keep client information out of personal AI accounts.

Everywhere. The safest design keeps client documents inside systems you control. A private assistant on your own documents answers from your files without sending them into a public chatbot, and our comparison of ChatGPT, Claude, Gemini, Grok and Muse sets out which plans train on your data and which don't. None of this is legal advice; check your own position with your professional body or adviser.

Making Tax Digital: more records, more often (UK)

Making Tax Digital for Income Tax is now live for sole traders and landlords with qualifying income over £50,000, who had to start from 6 April 2026. It extends to those over £30,000 from 6 April 2027 and over £20,000 from 6 April 2028. Clients in scope must keep digital records and send quarterly updates through compatible software.

For a practice, that turns a once-a-year scramble into four smaller ones. The jobs in the table above, chasing records and capturing receipts, stop being a seasonal headache and become a quarterly routine, which is exactly the kind of repeating, rule-based work automation handles well.

Will AI replace accountants?

No, and the profession agrees. In the Financial Cents survey, 90% said human judgement matters more as AI spreads, and only 19% trusted AI output enough to use it with limited review. What AI replaces is keying, chasing and first drafts. The accountants who benefit are the ones who hand those over and spend the time on advice, planning and the client conversations that technology can't have.

What we would do

For most small and mid-sized practices, we would start in this order:

1. Switch on your ledger's AI for categorising and matching, and US QuickBooks firms should plan the move to Intuit Accountant Suite before 31 December.
2. Automate record chasing from your practice management tool, because it saves the most time for the least risk.
3. Connect receipt capture to the ledger, with a review queue for anything unusual.
4. Set up answers to routine client questions before your busiest season: January in the US, or before the next quarterly deadline for UK clients under Making Tax Digital.

Each step is measured before the next one starts, so you always know what it is saving. If you are not sure where your firm stands, our AI consulting starts with exactly that question.